Construction sites are complex and dynamic environments, making navigation a challenge for robots. Imagine a robot trying to deliver materials amidst moving equipment and changing obstacles. Traditional methods like Simultaneous Localization and Mapping (SLAM) are computationally intensive. However, new research explores using Building Information Modeling (BIM) data combined with AI to create safer, more efficient robot pathfinding. This innovative approach leverages the rich spatial information within BIM to generate a virtual map of the construction site. Researchers then employ a Multi-Heuristic A* (MHA*) algorithm, enhanced with Artificial Potential Fields (APFs), to plan optimal paths. Think of it like creating a force field around obstacles, gently guiding the robot away from danger while still reaching its destination efficiently. To add another layer of intelligence, the researchers integrate Large Language Models (LLMs) like GPT-3.5. The LLM analyzes the textual descriptions of BIM elements (like "chair" or "grinder") to understand their potential danger. This allows the system to dynamically adjust the repulsive force fields, giving the robot a wider berth around hazardous objects. This combination of BIM, classical pathfinding algorithms, and the reasoning capabilities of LLMs results in significantly improved object avoidance. In tests, the AI-powered system enabled robots to maintain up to 80% greater distance from obstacles compared to traditional methods, all while maintaining comparable path lengths. This not only enhances safety but also boosts the trustworthiness of robots on construction sites by providing clear explanations for their navigation decisions. Future research aims to make this system even more adaptable by incorporating real-time obstacle avoidance techniques and dynamic adjustments based on changing site conditions. This research highlights the potential of combining BIM data with cutting-edge AI to unlock new levels of efficiency and safety in construction robotics.